PHYSICAL REVIEW E 67, 066501 共2003兲

Properties of the ultrashort gain length, self-amplified spontaneous emission free-electron laser
in the linear regime and saturation
A. Murokh,1 R. Agustsson,1 M. Babzien,2 I. Ben-Zvi,2 L. Bertolini,3 K. van Bibber,3 R. Carr,4 M. Cornacchia,4 P. Frigola,1
J. Hill,3 E. Johnson,2 L. Klaisner,4 G. Le Sage,3 M. Libkind,3 R. Malone,2 H-D. Nuhn,4 C. Pellegrini,1 S. Reiche,1
G. Rakowsky,2 J. Rosenzweig,1 R. Ruland,4 J. Skaritka,2 A. Toor,3 A. Tremaine,3 X. Wang,2 and V. Yakimenko2
1
UCLA Department of Physics and Astronomy, University of California at Los Angeles, 405 Hilgard Avenue,
Los Angeles, California 90095-1547, USA
2
Brookhaven National Laboratory, Upton, New York 11973, USA
3
Lawrence Livermore National Laboratory, Livermore, California 94551, USA
4
Stanford Linear Accelerator Center, Stanford, California 94309, USA
共Received 13 December 2001; published 4 June 2003兲
VISA 共Visible to Infrared SASE Amplifier兲 is a high-gain self-amplified spontaneous emission 共SASE兲
free-electron laser 共FEL兲, which achieved saturation at 840 nm within a single-pass 4-m undulator. The
experiment was performed at the Accelerator Test Facility at BNL, using a high brightness 70-MeV electron
beam. A gain length shorter than 18 cm has been obtained, yielding a total gain of 2⫻108 at saturation. The
FEL performance, including the spectral, angular, and statistical properties of SASE radiation, has been char-
acterized for different electron beam conditions. Results are compared to the three-dimensional SASE FEL
theory and start-to-end numerical simulations of the entire injector, transport, and FEL systems. An agreement
between simulations and experimental results has been obtained at an unprecedented level of detail.

DOI: 10.1103/PhysRevE.67.066501 PACS number共s兲: 41.60.Cr, 41.60.Ap, 41.85.Ja

I. INTRODUCTION beam microbunching measurements and nonlinear harmonics

generation are reported elsewhere 关17,18兴.
There is a great interest in the scientific community to-
wards development of the intense, short pulse, x-ray sources II. EXPERIMENTAL SETUP
to probe the dynamics of chemical and biological processes
The general layout of the experiment is shown in Fig. 1.
at the molecular and atomic levels. In one promising scheme,
The ATF 1.6-cell photocathode S-band gun generated a high
it was suggested 关1兴 to build an unseeded, single-pass, self-
brightness electron bunch, which was accelerated in two con-
amplified spontaneous emission free-electron laser 共SASE-
secutive SLAC-type linac structures to the VISA operating
FEL兲 based on the high brightness ultrarelativistic electron energy of around 71 MeV 关19兴. A 20° double-bend transport
beam presently available at SLAC 关2兴 and in the future at section served to direct the beam into Beamline III of the
TESLA 关3兴. The concept of the SASE-FEL was introduced in ATF experimental hall, where the VISA undulator was lo-
the early 1980s 关4,5兴, and in the following years the one- cated.
dimensional 共1D兲 theory has been extended to include three- The 4-m-long in-vacuum VISA undulator 关20兴 was built
dimensional 共3D兲 effects 关6,7兴, and to describe temporal and at SLAC as a Halbach array of 220 periods, each 1.8-cm
statistical properties of the SASE radiation 关8兴. long. The pole magnets provided an on-axis peak field of
In recent years, with the advancement of the photoinjector approximately 0.75 T. In addition, the undulator was
technology 关9兴, a series of successful experiments demon- equipped with a periodic quadrupole lattice, which generated
strated high gain SASE operation, in the IR 关10,11兴, visible a strong focusing gradient throughout the length of the un-
关12兴, and ultraviolet 关13兴 spectral ranges. Finally, a SASE-
FEL saturation has been obtained 关14兴 at 540 nm. The VISA
共Visible to Infrared SASE Amplifier兲 experiment reported
here was designed to further investigate the physical proper-
ties of the SASE-FEL relevant to the future LCLS operation.
In particular, the efficiency limits of the SASE process were
studied experimentally through optimization 关15兴 of the un-
dulator parameters for the best available low emittance pho-
toinjector beam produced at the Accelerator Test Facility
共ATF兲 关16兴. VISA achieved saturation at 840 nm within a
4-m-long undulator with the SASE power gain length of 17.9
cm—more than a factor of 3 shorter than what had been FIG. 1. Experimental layout of the ATF Beamline III: gun and
previously achieved in this spectral range. The focus of this linac area 共a兲; 20° double-bend dispersive section 共b兲; and VISA
manuscript is on the measured FEL performance at the fun- experimental area 共c兲, including the undulator and radiation diag-
damental wavelength; while other results such as electron nostics.

1063-651X/2003/67共6兲/066501共5兲/$20.00 67 066501-1 ©2003 The American Physical Society

MUROKH et al.
dulator, forcing the electron beam matched ␤ function to an
average value of about 30 cm. The permanent magnet pieces
were sorted to achieve field strength accuracy of better than
0.4%. The undulator structure was constructed out of four
individual sections, each 0.994-m long. During the magnetic
measurements performed at BNL 关21兴, the quadrupole mag-
nets were aligned and additional shimming was introduced to
provide a desired central trajectory straightness 关22兴 of better
than 80 ␮m rms throughout each pair of the consecutive
undulator sections. In order to align the magnetic axis of all
four individual sections to better than 20 ␮m throughout the
4-m length of the device an interferometric alignment proce-
dure was developed 关23兴 and implemented at the location of
the experiment. The undulator vacuum chamber was de-
signed to allow an independent support and control to each
undulator segment while in vacuum 关24兴; and a charge-
coupled-device 共CCD兲-based optical monitoring system was
developed 关25兴 to trace the relative positions of the undulator
sections with a 10-␮m accuracy throughout the pumping
cycles and if in-vacuum corrections were required.
The undulator and vacuum chamber design included eight
diagnostic ports 50-cm apart, starting 25 cm into the undu-
lator. The applications of these ports were both to measure
SASE radiation properties along the undulator length, and to
determine the electron beam position and envelope profile at
three points throughout every betatron period. The undulator
in-vacuum diagnostic probes were developed 关26兴 to inter-
cept the beam path inside the undulator during the measure-
ments. To ensure clearance through 3.6-mm openings in the
undulator frame the probes were constructed only 2.2-mm
wide, which is nonetheless many times both the electron and
radiation rms beam sizes. The active area of the probe—a
double-sided silicon mirror—could be actuated in two posi-
tions: to reflect a FEL radiation into the optical diagnostic
system, or to generate the optical transition radiation 共OTR兲
for the electron beam imaging system, respectively. The pol-
ishing of these miniature mirrors to the laser quality surface
finish presented a challenge, which motivated the use of sili-
con.
The initial design for electron beam diagnostic detectors
considered single crystal YAG:Ce 共where YAG is yttrium
aluminum garnet兲 scintillating screens. However, a diagnos-
tic test performed at the ATF indicated an intensity-
dependent blooming of the beam image on the YAG screen
with the electron beam parameters similar to the ones ex-
pected at VISA 关26兴. As an alternative, the OTR off the probe
mirror surface, coupled with polarizers and filters to reject
the SASE signal, was utilized to image the electron beam.
The relatively low intensity of the OTR, coupled with the
high optical noise environment inside the undulator, placed a
significant limitation on the dynamic range of the OTR-
based beam profile monitors 关27兴. Since beam trajectory con-
trol was one of the most technically challenging aspects of
the experiment, a very stable fiber coupled reference diode
laser was purchased and prealigned to the undulator mag-
netic axis with 20 ␮m accuracy, to provide a fiducial mark
for the electron beam centeroid trajectory measurements.
066501-2
PHYSICAL REVIEW E 67, 066501 共2003兲
TABLE I. Electron beam parameters and FEL performance for
the noncompressed 共case A兲 and compressed beam 共case B兲, respec-
tively.
Case A Case B
Electron beam energy 71.2 MeV 70.7 MeV
Beam charge, Q 250 pC 140 pC
Horizontal emittance, ␧ n 2.1⫾0.2 ␮ m 3.3⫾0.2 ␮ m
Peak current, I p 55 A 250 A
rms energy spread, ⌬␥/␥ ⬍0.10% 0.17%
FEL gain length, L g 29.7 cm 17.9 cm
SASE wavelength, ␭ r 831 nm 842 nm
No. of temporal spikes, M 4 –5 1–2
Total gain, G 7⫻104 2⫻108
III. ELECTRON BEAM MEASUREMENTS
Initial measurements of the electron beam properties were
performed at the exit of the linac 共Table I兲. Emittance was
determined with the quadrupole scanning technique, and op-
timized in the range of ␧ n ⬃1.9– 2.3 ␮ m. To determine the
current profile, a calibrated linear energy chirping was ap-
plied to the electron beam and the horizontal beam profile
was measured around the first bend of the dispersive section
of the beamline. For a sufficient amount of chirp, the hori-
zontal distribution was entirely dominated by the dispersion
in the bend, and thus could be converted into the beam cur-
rent profile. The measurement showed nearly flat top distri-
bution with a peak value of about 55 A.
To achieve the design goals, however, a much larger beam
current was required. This was achieved with the develop-
ment of a bunch compression mechanism, utilizing second
order effects in the dispersive section. Namely, the quadru-
poles tune in the dispersive section was chosen such that the
transport coefficient T 566 was large and negative. As a result,
the specific amount of the linac phase detuning resulted in
the two effects: 共1兲 the beam acquired a linear energy chirp-
ing in the linac 共⌬␥/␥兲; and 共2兲 the initially small compres-
sion coefficient R 56⬇T 566(⌬p/p) increased to a significant
value, as the beam centroid momentum was decreased by
⌬p. The combined effect, given the proper choice of the
linac phase detuning and the beamline tune T 566 values, re-
sulted in a strong longitudinal bunch compression, and a
much larger longitudinal current 关28兴.
To measure the compression effect, a coherent transition
radiation 共CTR兲 diagnostics was implemented after the dis-
persive section of the beamline. The integrated intensity in
the coherent fraction of the transition radiation spectrum
generated by the electron bunch has a strong dependence on
the bunch length ␴ ␶ 关29兴. To measure the change of CTR
intensity as a function of the linac phase detuning, the signal,
generated by a 45° mirror intersecting the beam path after
the dispersive section, was directed into a far-infrared bo-
lometer. The measured CTR intensity 共Fig. 2兲 showed a
strong correlation with the linac rf phase, and thus allowed
us to verify the compression process. In addition, an insert-
able low-pass filter with a known frequency roll-off was used
for quantitative assessment of the high frequency CTR com-
PROPERTIES OF THE ULTRASHORT GAIN LENGTH, . . .
FIG. 2. Measured CTR signal intensity exhibits a sharp peak
within a narrow window 共⬃1.5°兲 of the linac RF phase.
ponent. Measurements with and without the filter provided a
benchmark for the numerical modeling of the system. As a
result it was found that at the optimized value of the linac rf
phase the bunch is compressed longitudinally by a factor of
5, yielding a peak current of 250 A.
IV. FEL MEASUREMENTS AND ANALYSIS
Without the bunch compression, VISA lased at 830 nm,
with a SASE pulse energy of 5 nJ measured at the undulator
exit, corresponding to a gain length of about 30 cm 共Table I,
case A兲. Spectral measurements 关Fig. 共3a兲兴 showed a typical
spiky structure, indicating the superposition of many tempo-
ral modes 共spikes兲, as expected for a lasing core of the elec-
tron beam about 500-␮m long 关27兴. With the compression in
place 关Fig. 3共b兲兴, the spectrum was wider and lacked any
structure, suggesting the single temporal mode 关30兴, consis-
tent with the shorter bunch length. With the compressed
beam the SASE energy at the undulator exit increased by
many orders of magnitude reaching up to 20 ␮J and showing
saturation when the optimized operating point was found.
Since the SASE process develops from a random shot
noise in the electron longitudinal distribution, it was impor-
tant to analyze the statistical properties of the FEL process,
such as shot-averaged performance, and shot-to-shot fluctua-
tions of the SASE intensity. However, that was not a trivial
task since the bunch compression factor, being a strong func-
tion of the linac rf phase, was changing on the fast time
scale, following the jitter in the rf system. As a result, the
measured SASE fluctuations were strongly dominated by the
linac rf noise. To overcome this problem, a collimator in the
dispersive section was used to monitor the linac phase evo-
lution by measuring the charge loss in the collimator. The
nonlasing tale of the dispersion-dominated beam was clipped
by the collimator, and the amount of clipping was monitored,
allowing us to select for data analysis only those shots that
were clipped by the same amount 关26兴.
The shot-to-shot distribution of the SASE process de-
pends on the number of independent lasing modes contribut-
ing to each shot and characterized with the ⌫-function distri-
bution 关31兴. In VISA the electron beam transverse size was
typically only 60–90-␮m rms, so that at the lasing wave-
length the Raleigh range was much smaller than the gain
066501-3
PHYSICAL REVIEW E 67, 066501 共2003兲
FIG. 3. Comparison of the SASE radiation spectra measured
without 共A兲 and with 共B兲 compression. The spiky structure in case
A indicates many longitudinal modes 共i.e., a longer bunch length兲.
length, and full transverse coherence was established early.
As a result, the statistical properties of the intensity distribu-
tion were related entirely to the number of longitudinal
spikes in the SASE radiation. In the absence of compression
关Fig. 4共a兲兴, a ⌫-function fit indicated 4 –5 temporal modes,
and with the compressed beam, a distribution inside the un-
dulator 关Fig. 4共b兲兴 resembled a negative exponential, charac-
teristic of a single spike lasing 关both distributions are consis-
tent with the spectra in Figs. 3共a兲 and 3共b兲, respectively兴. At
saturation, however, the SASE intensity distribution from the
compressed beam has changed dramatically 关Fig. 4共c兲兴, since
it was no longer dominated by the shot noise fluctuations, but
rather by the shot-to-shot fluctuations in the saturation
length, which can be as large as few gain lengths for short
bunches 关30兴.
With the charge selection criterion described above, one
could compare data sets obtained at the different diagnostic
ports along the undulator, while the system stayed at the
same operating point. As a result, the FEL gain evolution
along the undulator length was studied. The measured inten-
sity of SASE radiation had noticeably exceeded the level of
a broadband spontaneous emission starting from the diagnos-
tic port 3 共1.25 m into the undulator兲. In the curve shown in
Fig. 5, the fit to the exponential part of the curve yielded a
power gain length of 17.9 cm, corresponding to 3D FEL
parameter ˜␳ ⬇0.0046, reduced from the 1D limit by a factor
FIG. 4. SASE fluctuations 共a兲 without compression and at high
gain measured 共b兲 inside the undulator and 共c兲 at the undulator exit.
MUROKH et al. PHYSICAL REVIEW E 67, 066501 共2003兲

of (1⫹⌳)⬇1.75, mostly due to diffraction. Near the undu-

lator exit, the curve deviated from the exponential behavior,
indicating saturation, which was quite consistent with the
theoretical estimate of the saturation length 关6兴, L sat ⬇␭ u /˜␳
⫽3.9 m.
The measured bandwidth at saturation, (⌬ ␻ / ␻ ) f ⬇1.2%,
was also found in agreement with the value predicted from
theory, if one would add the contribution from the beam
energy spread term: ⌬ ␻ / ␻ ⬇2˜␳ 关 1⫹2(⌬ ␥ / ␥˜␳ ) 2 兴 . Finally,
the SASE power at saturation was estimated to be about 75
MW based on the SASE energy and bandwidth measure-
ments, which was consistent with the 3D scaling law P sat
⬃1.6P beam˜␳ /(1⫹⌳) 关32兴. The final state SASE gain G
⬅(d P/d ␻ ) f /(d P/d ␻ ) noise was evaluated using the treat-
ment given in 关33兴, using only the measured parameters:
FIG. 5. Measured SASE evolution along the undulator length

冉 冊 冉 冊
and numerical simulations 共gray lines are the rms boundaries of the
3 ␲ K 2 关 JJ 兴 2 I b m e c 2 ⌬ ␻ L̃ g set of GENESIS runs兲. The amplification curve yields power gain
G ⫺1 ⬇ .
4 1⫹K 2 /2 I A P f /k r c ␻ ␭
f u
lengths of 17.9 cm and saturates near the undulator exit.

With the undulator parameter K⫽1.26, and the average
beam current over the spike length I b ⫽210 A, the gain value
at saturation was found to be G⬇2⫻108 .
V. NUMERICAL MODEL OF THE EXPERIMENT
Since the details of the compression mechanism during
the beam transport were critical for understanding the elec-
tron beam properties inside the undulator, a full cathode-to-
undulator simulation of the experiment was needed to model
the SASE results. Starting from the photoemission at the
cathode, a UCLA version of PARMELA 关34兴 was used to
model acceleration and emittance compensation through the
gun and linac sections. The PARMELA results matched the
beam properties measured after the linac 共Table I兲. The code
ELEGANT 关35兴 was used to model the beam dynamics and the
bunch longitudinal compression in the double-bend disper-
sive section, including the effects of coherent synchrotron
radiation 共CSR兲. The beam compression was studied as a
function of linac rf phase detuning, and the ELEGANT output
should be tested against the CTR data 共Fig. 2兲 to achieve a
good agreement between a model and the experimental data
over a wide range of system parameters. One critical result
of the numerical simulations was understanding of the emit-
tance growth mechanisms in the dispersive section. The EL-
EGANT CSR model showed that the slice horizontal emit-
tance increase due to CSR was not significant, i.e., ⌬␧ CSR
n radiation were measured and found to be in good agreement
⬇0.3 ␮ m rad. However, when the beam energy offset was
set to model the maximum compression case, the noncom-
pensated dispersion term, ␩ ⬃5 cm, combined with the large
slice energy spread in the compressed beam, dominated the
emittance growth, bringing the effective horizontal slice
n ⬇3.5 ␮ m rad.
emittance inside the undulator up to ␧ eff
The 3D time-dependent code GENESIS 1.3 关36兴 was used
to model the radiation evolution inside the undulator. An
ELEGANT output file was converted into an input file for
GENESIS preserving all 6D properties of the electron beam
phase space 关37兴. The simulations of the FEL process are
shown in Fig. 5 together with the experimental data. The
measured SASE gain curve was well within the standard
deviation of a set of GENESIS runs with randomly varying
shot noise. In addition to a good agreement with measured
SASE gain evolution, the developed numerical model accu-
rately reproduced spectral and angular properties of the ra-
diation. Figure 6 shows the hollow angular distribution of the
SASE intensity measured at the exit of the VISA undulator,
together with the one generated from the far field GENESIS
output. Such an unusual SASE angular profile was repeat-
edly observed in the experimental measurements. It was re-
produced numerically, when the strongly asymmetric and en-
ergy correlated horizontal phase space distribution in the
electron beam, originated in the bunch compression process,
was carried from the ELEGANT output into the FEL simula-
tions. Having this unusual result reproduced with GENESIS
indicated a new level of insight into the dynamics of the
electron beam and SASE FEL systems, obtained by the syn-
thesis of copious diagnostic measurements and rigorous, de-
tailed simulations.
In conclusion, the VISA experiment presented an exten-
sive study of the dynamics of a high gain SASE FEL driven
by a high brightness, longitudinally compressed electron
beam in a strong focusing, compact undulator. A gain of 2
⫻108 and saturation were obtained, and the statistical prop-
erties of the SASE signal at saturation were measured. SASE
evolution as well as spectral and statistical properties of the
FIG. 6. Far field angular distribution of the SASE radiation 共a兲
measured on the CCD camera and 共b兲 simulated with GENESIS.

066501-4
PROPERTIES OF THE ULTRASHORT GAIN LENGTH, . . .
with the 3D FEL theory. To explain fine details in the per-
formance of the system, it was necessary to investigate, both
experimentally and through numerical modeling, the dynam-
ics of the electron beam throughout the entire injector-
transport-FEL process. The developed start-to-end integrated
computational model showed an excellent agreement with
the measurements, and gave strong insights into the physical
关1兴 C. Pellegrini, Proceedings of the Workshop on 4th Generation
Light Sources, SLAC, 1992 共unpublished兲.
关2兴 Linear Coherent Light Source Design Study Report No.
SLAC-R-521, 1998 共unpublished兲.
关3兴 Linear Collider Conceptual Design Report No. DESY 97-48,
1997 共unpublished兲.
关4兴 R. Bonifacio, C. Pellegrini, and L. Narducci, Opt. Commun.
50, 373 共1984兲.
关5兴 A. M. Kondratenko, E. L. Saldin, Part. Accel. 10, 207 共1980兲.
关6兴 K. J. Kim, Phys. Rev. Lett. 57, 1871 共1986兲.
关7兴 S. Krinsky and L. H. Yu, Phys. Rev. A 35, 3406 共1986兲.
关8兴 R. Bonifacio et al., Phys. Rev. Lett. 73, 70 共1994兲.
关9兴 J. S. Fraser et al., IEEE Trans. Nucl. Sci. 32, 1791 共1985兲.
关10兴 M. Hogan et al., Phys. Rev. Lett. 80, 289 共1998兲; 81, 4867
共1998兲; D. C. Nguyen et al., ibid. 81, 810 共1998兲.
关11兴 L. H. Yu et al., Nucl. Instrum. Methods Phys. Res. A 445, 301
共1999兲; Proceedings of the FEL Conference 2000, Duke Uni-
versity 共2000兲.
关12兴 M. Babzien et al., Phys. Rev. E 57, 6093 共1998兲.
关13兴 J. Andruszkow et al., Phys. Rev. Lett. 85, 3825 共2000兲; V.
Ayvasyan et al., ibid. 88, 104802 共2002兲.
关14兴 S. V. Milton et al., Phys. Rev. Lett. 85, 988 共2000兲; Science
292, 2037 共2001兲.
关15兴 H.-D. Nuhn 共private communication兲.
关16兴 X. Wang, X. Qiu, and I. Ben-Zvi, Phys. Rev. E 54, 3121
共1996兲.
关17兴 A. Tremaine et al., Phys. Rev. Lett. 88, 204801 共2002兲.
关18兴 A. Tremaine et al., Phys. Rev. E 66, 036503 共2002兲.
关19兴 X. Wang et al., PAC 1999 Proceedings, edited by A. Luccio
and W. MacKay 共IEEE, Piscataway, 1999兲, p. 3495; V. Yaki-
066501-5
PHYSICAL REVIEW E 67, 066501 共2003兲
mechanisms of the complex phenomena present in the ex-
perimental system.
ACKNOWLEDGMENT
This work was done under DOE Contract No. DE-FG03-
92ER40693.
menko et al., FEL 2001 Proceedings 共unpublished兲, p. 277.
关20兴 R. Carr et al., Phys. Rev. ST Accel. Beams 4, 122402 共2001兲.
关21兴 G. Rakowsky et al., PAC 1999 Proceedings, edited by A.
Luccio and W. MacKay 共IEEE, Piscataway, 1999兲, p. 2698.
关22兴 P. Emma and H. D. Nuhn, FEL 1998 Proceedings 共Elsevier
Science B.V., Amsterdam, 2000兲, Vol. II, p. 35.
关23兴 R. Ruland et al., PAC 1999 Proceedings, edited by a. Luccio
and W. MacKay 共IEEE, Piscataway, 1999兲, p. 1390.
关24兴 M. Libkind et al., PAC 1999 Proceedings, edited by A. Luccio
and W. MacKay 共IEEE, Piscataway, 1999兲, p. 2477.
关25兴 A. Tremaine, A. Murokh, and X. Wang, BNL Report No.
68170, 2001 共unpublished兲.
关26兴 A. Murokh et al., PAC 2001 Proceedings, edited by P. W. Lu-
cas and S. Webber 共IEEE, Piscataway, 2001兲, pp. 2748 and
1331.
关27兴 A. Murokh, Ph.D. thesis, UCLA, 2002.
关28兴 J. B. Rosenzweig 共private communication兲.
关29兴 A. Murokh et al., Nucl. Instrum. Methods Phys. Res. A 410,
452 共1998兲.
关30兴 R. Bonifacio et al., Phys. Rev. Lett. 73, 70 共1994兲.
关31兴 E. L. Saldin et al., Opt. Commun. 148, 383 共1998兲.
关32兴 M. Xie, PAC 95 Proceedings 共unpublished兲, p. 183.
关33兴 L. H. Yu and S. Krinsky, Nucl. Instrum. Methods Phys. Res. A
285, 119 共1989兲.
关34兴 E. R. Colby, Ph.D. thesis, UCLA, 1997.
关35兴 M. Borland, Advanced Photon Source Report No. LS 287,
2000 共unpublished兲.
关36兴 S. Reiche, Nucl. Instrum. Methods Phys. Res. A 429, 243
共1999兲.
关37兴 S. Reiche 共private communication兲.